Andriy Viy

I build Apps, Websites, AI agents and automation for small businesses 🧑🏻‍🔬

Fluent in 🇩🇪🇬🇧🇵🇱🇺🇦

📍🇮🇹

[Andriy Viy @andriyviy]

@christiannonis I am using omp it’s like pi but on steroids
So it’s slow to adapt on recent model releases so I was confused OpenAI wasn’t allowing it to use with 3rd party harnesses

https://t.co/E7FgtOel5x

[Andriy Viy @andriyviy]

I can’t use GPT Luna 6 in oh my pi anymore ? Whats going on???

[Andriy Viy @andriyviy]

@RES_YutoR I tried to set it and got this:

[Andriy Viy @andriyviy]

@benglickenhaus @willcb Do Luna on max and wait all day for frontier results 😇

[Andriy Viy @andriyviy]

@Yuchenj_UW I am team Opus 5.5 Med now

[Andriy Viy @andriyviy]

So it’s cheaper to run Opus 5.5 on High than Astra 6 on Max reasoning... And it is also smarter. Okay, next month I am switching back to Claude.

[Andriy Viy @andriyviy]

@kamilbaranek_ 0

[Andriy Viy @andriyviy]

I hope Sol 6 is less retarded than 5.6 😮‍💨

[OpenAI @OpenAI]

Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.

GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.

We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.

[Andriy Viy @andriyviy]

@christiannonis @cursor_ai Oh boy 🥲 I barely survive with my 100$ plan on regular spec driven dev.
By I liked your idea of building something based on OpenAI’s Symphony.
But I am afraid I will be the bottleneck for such workflow (for now)😆

[Andriy Viy @andriyviy]

@0xPrajwal_ @jeffersonmendza Chain debit to credit with a then that has a valid response, else catch and don’t credit

[Andriy Viy @andriyviy]

@sayodotfun Rocking S tier here 😤

[Andriy Viy @andriyviy]

@neural_avb Is it good for anything really or it’s all hype?

[Andriy Viy @andriyviy]

@ibocodes Maybe for email auto sorting
Tried to use it for browser control but it’s not good for it despite the hype

[Andriy Viy @andriyviy]

Today I’ll be starting 50 Jev agent loops on this bad boy:

[Andriy Viy @andriyviy]

@RoseSilicon Idk I build stuff for myself anyway, useless as a buyer though 😁

[Andriy Viy @andriyviy]

@RohOnChain … and loose all your live savings

[Andriy Viy @andriyviy]

@typesafeai ✨rate limited✨

[Andriy Viy @andriyviy]

@StevenDarlow @typesafeai @Teknium Bro why you put skills in a zip? Does it actually work or it’s just bait?

[Andriy Viy @andriyviy]

So I am still sticking with surf-cli, even though it’s freaking expensive to use with frontier LLM models, but at least it can reason about what’s going on.

Repos used: browser-use/jev-ultrafast and nicobailon/surf-cli.

[Andriy Viy @andriyviy]

I wanted to test the Jev hype.
For predictable interactions, it is genuinely fast, like selecting some filters in Airbnb.

But the. I asked it to create a three-page Google Form with six questions, publish it, fill it out and submit it.

[Andriy Viy @andriyviy]

But in this test, getting a finished result still required a lot of work around the model. I couldn't just hand it the task and walk away.